1996 FatigueFracturePyrolyticCarbon Ritchie

Fatigue and Fracture of Pyrolytic Carbon: A Damage-Tolerant Approach to Structural Integrity and Life Prediction in “Ceramic” Heart Valve Prostheses Robert O. Ritchie Department of Materials Science and Mineral Engineering, University of California, California, USA Background and aims of the study: The fracture and fatigue properties of Si-alloyed LTI pyrolytic carbon, and pyrolytic carbon-coated graphite are described as a framework for establishing damage-tolerant analyses for maintaining structural integrity and for pre- ting the lifetimes of mechanical heart valve prostheses fabricated from these materials. ‘Methods: The analyses are based on fracture-mechanics concepts and provide conservative (worst-case) estimates of the time, or number of loading cycles, before the valve will fail, or more precisely for pre existing defects in valve components to grow subcritically to critical size under elevated physiologic loading and environmental con Mechanical heart valve prostheses are designed to regulate blood flow continuously in hostile physiologic environments for periods in excess of patient lifetimes. Valve components are exposed to cyclic loading, flexing and bending, wear at surfaces exposed to artic- ulation, and cavitation erosion on surfaces exposed to blood flow. Silicon-alloyed low-temperature isotropic (LTD pyrolytic carbon has been shown to be a success- ful material for this application (1-4), has been used in ‘over 600,000 implanted valves and is currently used in the majority of artificial heart valves that are commercially available (5). Most modern heart valves are generally constructed with two semi-circular leaflets (bileaflets), or with a tit- ing disc (occluder), which opens) and close(s) as the heart beats, thereby regulating blood flow under near- normal rheological conditions. The leaflets or dise are contained in a circular housing (or orifice), which in certain valves is stiffened by a metallic restraining ring; a cloth sewing ring is affixed to the outer diameter to facilitate attachment to cardiac tissue. In the US, the disc or leaflets are invariably manufactured from pyrolytic carbon-coated graphite; the iamaiecrapmeee ake. tach! Deparment of Motels Since and Mine Erpntening Unis oCliiomia Berkley Ch SIa0 700A Results: For structural life in excess of patient lifetimes, a minimum required detectable defect size is computed which must be detected by quality-control procedures prior to the device entering service; this defect size is typically of the order of tens of microns for such “ceramic” valves, compared to sizes in the hundreds of microns for corresponding metal valves. Conclusions: It is concluded that in light of the brittle nature of pyrolytic carbon and the unacceptable cost of mechanical valve failures, the use of such analyses should be regarded as essential in order to provide ‘maximum assurance of patient safety. ‘The Journal of Heart Valve Disease 1996;5Suppl. D: 89-831 housing ean also be manufactured from this material, oralternatively from pure pyrolytic carbon, titanium or cobalt-chromium alloys. Unfortunately, the complex shapes of some of the metallic components have required the use of casting and welding technologies in their fabrication. As a result, serious structural prob- lems have arisen with certain of these devices, leading, to many implant failures due to fatigue cracking in the valve housing (6-9). Accordingly, the majority of heart valves implanted in the US. today are made entirely from ceramictike material, ie, from pyrolytic carbon and/or a pyrolytic carbon/graphite laminate. Exam- ples of such valves are shown in Figure 1 One of the primary restrictions on artificial heart valves is their potentially uncertain lifetime under complex physiologic loading due to degradation by cyclic fatigue or stress corrosion. While hundreds of thousands of pyrocarbon heart valves have been successfully implanted, over forty recent structural failures, involving principally leaflet fractures, have been reported and in many cases attributed to cyclic fatigue (10). An example of such an in vivo leaflet failure is shown in Figure 2. In this instance, cracking was pos bly initiated by cyctic fatigue prior to catastrophic overload fracture. Although the latter conclusion may be somewhat questionable due to the difficulty in distin- guishing fractographic features produced under cyctic © Copyright by ICR Publishers 1996S10 Fatigue and fracture of pyrolytic carbon R.O. Ritchie Figure 1: A selection of commercial pyrolytc-carbon bilenflet ‘ad tilting-dise prosthetic heart valoes. A metalic tilting-iise talve (with a Co-Cr alloy housing and a pyrolytic carbon/graphite occlude) is shown for comparison (bottom left versus impact loading conditions (11), these failures do serve to illustrate that careful mechanical design and accurate life prediction are essential elements for the safe, reliable operation of heart valve prostheses, which must endure sustained physiologic loading and environmentally induced degradation throughout their service lifetime. Since the human heart beats some 40 million times per year, the primary mechanism of structural degradation can be considered to be cyclic fatigue. Accord ingly, to maintain structural integrity, prosthetic heart valves must be designed to endure fatigue lifetimes in excess of 10” cycles in physiologic environments. To achieve this, fatigue-life estimation procedures have been developed (12), initially for metal (12) and subse- quently for pyrocarbon (13) valves, based on both stress/life and so-called damage tolerant approaches; currently, however, because of the safety-critical nature of replacement heart valve applications, Food and Drug Administration (FDA) requirements (14,15) demand that for mechanical valves, damage-tolerant design analyses are used. ‘The latter approach relies on the fracture-mechanics based concept that a conservative (worst case) estimate of the structural life can be defined in terms of the time, ‘or number of loading cycles, for the largest undetected crack to grow to critical size, generally defined as the “onset of catastrophic failure of the component in terms, of the material's fracture toughness, K.. A similar analysis might be required for other forms of subcritical crack growth, for example, by stress-corrosion cracking (static fatigue) in components subject to sustained (non-cyclic) loads. Im this paper, damage-tolerant design and life-pre- J Heatt Valve Dis 8. Suppl. Figure 2: Example of as in vivo failure ina pyrolytic-carbon canted graphite leaflet, where cracking ions possibly initiated in eycie fatigue prior to final averlod fracture, diction procedures are described for mechanical heart- valve prostheses, with specific reference to cardiac devices manufactured from ceramic-like pyrocarbon materials. The analyses are performed on a hypotheti- cal bileaflet valve and serve to determine the safe implant life as a function of the size and nature of pre- existing defects in the valve material. Based! on such calculations, a minimum required detectable crack size (22) is computed, which must be discovered by quality control procedures prior to the device entering service to ensure safe operation of the valve. It is shown that this minimum erack size can be of the order of tens of microns for pyrocarbon valves, compared to sizes in the hundreds of microns for the metal valves. Pyrolytic carbon materials Structure and processing ‘Three types of carbon are commonly used for biomedical devices; the low-temperature isotropic (LTD form of pyrolytic carbon, glassy (vitreous) carbon, and the ultra low-temperature isotropic (ULTD form of vapor-deposited carbon; all have a disordered lattice structure and are collectively referred to as turbostratic carbons (1,24). Although pyrolytic carbons were originally developed for elevated-temperature applications (eg., coatings for nuclear fuel particles), LT pyrolytic carbon has found greatest appeal in the biomaterials industry for mechanical cardiac valve components as it has been shown to be highly thrombore- sistant and to have inherent cellular biocompatibility with blood and soft tissue (4); moreover, it displays good durability, strength and resistance to wear (16,17), and had been thought to be immune to cyclic fatigue failure (18,19)| Heart Valve Dis Vol 5 Suppl. ) © Figure 3: The crystallographic arrangement of carbon atoms in (a) hexagonal graphite rohere parallel layer planes are in a regular sequence, and (b)« turbostratic carbon where the lay sare arranged without order. Randomly oriented turbostea ic crystallites are assembled f0 produce a bulk material in (c). Adapted from reference 3. ‘The structure of turbostratic carbons is closely relat- ced to that of graphite (Fig, 3a). In a single crystal of graphite, the carbon atoms are covalently bonded in planar hexagonal arrays which are stacked in layers in a regular ABAB sequence with relatively weak van der ‘Waals interlayer bonding; in polycrystalline graphite, these crystals, or “erystallites”, typically have diame- ters larger than 100 nm. With turbostratic carbons, the stacking sequence is disordered through random rota- tions or displacements of the layers relative to each other (Fig. 3); moreover, wrinkles or distortions may exist within each layer from imperfect matching of small layer segments due to missing carbon atoms. This structural distortion has a marked effect on dens ty, which can vary from 1,400 kgm? to a theoretical limiting value of 2,200 kgm. Crystallites in tur- bostratie carbon are much Smaller (10 nm) than in graphite, and are randomly oriented to produce the Fatigue and fracture of pyrolytic carbon $11 RO. Ritchie bulk material (Fig. 3e) (20); thus, on a macroscale which is large compared to the crystallite size, the material behaves in an isotropic fashion. For heart valve devices, a silicon-alloyed LTI- pyrolytic carbon is used, either as a =250-um thick coating on a polycrystalline graphite substrate or as a monolithic material where the substrate is machined away; 10 to 20 wt% silicon is added to improve mechanical properties, ic, stiffness, hardness, wear resistance, without significant loss in biocompatibility. Components are typically made by co-depositing, car- ‘bon and silicon carbide on the graphite substrate via a chemical vapor-deposition, fluidized-bed process using a gaseous mixture of a Si-containing carrier gas with a hydrocarbon (e., propane, methyltrichlorosi- lane and helium gas mixtures); the pyrolysis occurs at temperatures of 1000°C to 1500°C (2D. The resulting, pyrocarbon contains typically 10 wt% silicon, often in the form of discrete sub-micron B-SiC particles, randomly dispersed in a matrix of roughly spherical micron-size “growth features” (subgrains) of suberys talline turbostratic carbon, containing =10 nm sized -ystallites (22,23). Residual stress Biomedical turbostratic carbon coatings often exist in 4a state of residual stress. Like most coatings cooled from a higher processing temperature, the residual stress state depends on the difference in thermal expansion mismatch between the coating and the underlying substrate material. In addition, variation in the struc ture and grain size of the coating from the interface with the substrate to the coating surface may also affect the residual stress state. Both these effects are also dependent on the coating thickness. For LTI pyrolytic carbon-coated graphite components, the greater ther- ‘mal expansion coefficient of pyrolytic carbon compared to graphite results in a tensile residual stress in the coating. Depending on processing temperature and microstructural variation, various measurements seem to indicate a tensile stress of up to 60 MPa, which is often at a maximum near the surface of the coating and approaches zero at the pyrolytic carbon graphite interface (13), Such residual stresses may be particularly detrimental to the integrity of the coating as their value must be added to the nominal applied stress when ‘evaluating the likelihood of fracture in the coating, Mechanical properties ‘The mechanical properties of turbostratic carbons are closely related to their coating density as the extent of porosity affects the internal area over which stress is, distributed; properties also depend on specific compo- sition and microstructure, inclucling the volume frac- tion of SiC particles and the size of the crystallites and$12. Fatigue and fracture of pyrolytic carbon R.O. Ritchie t aia y race - igure 4: Experimental techniques used to measure cyclic fatigue-crack propagation rats, showing (a) half sound compact (lenfet) specimen and procedures used to monitor crack length for long trough:-thickness (edge) cracks, and (b)eantilever-bend specimen (with micro-indents) showing semi-elliptcal surface erack configuatiox for tests om sina eracks (26). “growth features” (2,16,17,24). Unlike certain forms of carbon with a high degree of preferred orientation and high density which are extremely strong, andl stiff in directions parallel to the preferred layer planes, because crystallites are randomly arranged and the ‘material is not fully dense, LTT pyrolytic carbon has an tuncharacteristicaly low modulus and comparatively high strength. Values of Young’s modulus are typically between 27 and 31 GPa, which is close to the upper bound value for bone of 21 GPa and an order of magnitude less than that of surgical stainless steel or tita um alloys. The hardness, and hence wear resistance, of pyrolytic carbons is also a function of crystallite size and SiC content; in fact, Si-alloyed pyrolytic carbon has superior wear resistance to either graphite or unal- loyed pyrolytic carbon 2). ‘The high strength and low moduli of turbostratic carbons results in large strains to failure when compared to other brittle ceramics, ie,, LTI pyrolytic carbon has a fracture strain of 1.5%- 2.0%, compared to Polycrystalline Property Graphite Substrate Density (kgm 1500-1800 Crystallite Size (nm) 15-250 Expansion Coefficient (10 K*) 055 Hardness (DPH) 50-120 ‘Young's Modulus (GPa) 412 Flexural Strength (MPa) 65-300 Fracture Strain (2) 04-07 Fracture Toughness, K,(MPavm) “AS | Heart Valve Dis Vol.5.Suppl.1 0.1% for alumina and 041%-0.7% for polycrystalline graphite. This is thought to result in part from. the network of strong covalent C-C bonds jn. the graphitic layers which must be broken before failure by shear or cleavage can occur. However, fracture toughness values, K, measured in the presence of a pre-existing. crack, are low compared to Vy 25,26), ‘The toughness of pure Si- alloyed LTI pyrolytic car bon or pyrolytic carbon coated graphite is of the order of 1 to 2 MPavm, ie,, only slightly higher than that of soda lime glass and small compared toa K, of =3 to MPaym for alumina and =3 to 15 MPavm for nia). A list of structural and mechanical properties for polycrystalline graphite, Si-alloyed LTI pyrolytic carbon and pyrolytic-carbon coated graphite is given in Table Biocompatibility High-purity turbostratic carbons, like LTI pyrolytic carbon, have exceptionally good cellular biocompatibility and thromboresistance @). For cardiovascular applications, compatibility with blood has received the most attention; unlike soft and hard tissue reactions to implanted materials which develop slowly, the rejec- tive reactions in blood are dramatic and swift. Most ‘materials in contact with blood quickly activate the tissue’s clotting mechanism; LT carbon has been shown to be equivalent to siliconized glass, which causes little Tie Srworal and medanizal proper of graphite St Aye pyri abo, wd graphitic carbon ania) Si-lloyed LT Pyrolytic-Carbon Coated Graphite 1500-2200, Si-Alloyed LT Pyrolytic Carbon 1700-2200 26 232-412 345, 1520 10-26| Heart Valve Dis Vol. 5. Suppl | damage to blood. Theories to explain the excellent cellular biocompatibitily of LTI carbon with blood range from conditioning of the carbon surface with a passi- vating protein layer through selective adsorption to the complete inertness of the LTI carbon surface to proteins in general. A general observation, however, is that smooth or polished surfaces are better than rough surfaces, possibly because roughness may provide locations where cells can adhere and serve as thrombotic, nuclei. Fatigue and fracture mechanics ‘The prediction of a safe lifetime is clearly a vital step in the design and operation of any structural component. In situations where stresses are primarily alternating, as in the case of heart valve prostheses, there are two approaches that can be used, namely the classical stress/life (S/N) approach and the damage- (or defect) tolerant approach which is based on fracture mechanics (12). With $/N analyses, smooth-sided or notched test specimens are cycled to failure at specific values of the stress or strain (defined as both a range and a mean) with the objective of relating the total life (ie., the number of cycles both to initiate andl propagate a crack to failure) to the stress or strain (27) Lifetime predictions are made based on such S/N. data, often following adjustments for such variables as notch geometry, variable-amplitude loading, environment, frequency, and so forth. Although used for numerous engineering situations, such as in the design (of small components such as gears, the approach is not favored for safety-critical situations, such as prosthetic heart valves, where loss of life is a possibility. This is because lifetimes are computed from laboratory- derived data representing the time or cycles both to initiate and propagate a crack; since actual components, often contain significant pre-existing cracks, there is always a potential for non-conservative lifetime prediction with the S/N approach. Accordingly, where the cost (either social or finan- cial) of a failure is too high, the alternative damage-tolerant approach must be adopted. This approach essentially assumes that all components already contain pre-existing defects and life is determined in terms of how long it takes these defects to extend to failure; such an approach has become mandatory for many nuclear and aerospace applications. Specifically, the concept of damage-tolerant design relies on the notion that the minimum safe structural life of a component can be determined from the time, or number of loacing cycles, for incipient cracks to grow subcritically to instability or until catastrophic failure occurs. For a “ceramic” heart valve prosthesis subjected to roughly 40 million cycles a year, this translates into the Fatigue and fracture of pyrolytic carbon § 13 RO. Ritchie time for the largest pre-existing defect to grow by cyclic fatigue (and/or stress corrosion) until final fracture ensues at K, (12). Accordingly, to estimate this life, a characterization of subcritical crack-growth rales for laboratory samples or actual components tested under realistic service (Le, in vitro) conditions must be deter mined, and then integrated between the limits of the nd final crack sizes. Such characterization can be achieved using fracture mechanics, the principles of which are described below. Fracture mechanics Concepts For small scale yielding conditions where the overall deformation in the component is nominally linear elastic and only limited inelastcity (e.g. plasticity) is present, the “driving force” for crack extension can be defined in terms of a stress intensity K which uniquely and autonomously characterizes the distribution of local stresses and displacements in the vicinity of the crack tip, viz: K = Qo Va, o where a is the crack length, dy» is the applied stress (which can inchude both externally applied stresses and residual stresses), and Qis a geometry factor (which is generally a function of a) of order unity. Sinee K characterizes the local crack-tp fel, its value can therefore be correlated to the extent of crack extension, Such concepts can be readily used to quantify the toughness of the material, specifically in terms of the critical value ofthe stress intensity atthe onset of unsta~ ble (i.e, catastrophic) fracture. Provided conditions of small scale yielding and plane strain exist, this critical value of stress intensity is referred to as the fracture toughness, K,, and is considered to be a material constant, Conditions of small scale yielding and plane strain are relevant provided the extent ofthe loca inelasticity, ‘eg, the crack-tip plastic-zone size, is small compared to the in-plane (e.g. crack length a) and out-of-plane (e.g test-pieee thickness B) specimen directions, respectively. ‘According to ASTM. standards (28), this can be achieved when 1,B 22.5 (K/o,?, where G, is the yield or flow stress ofthe material The fracture totighness can be used to define another important concept, that of the critical erack size, which represents the largest crack that can be tolerated by the component without cata- trophic failure. Since failure occurs at K = K, the critical crack size, for a particular applied stress, can be deter~ mined simply from the relevant K solution, e.g, from Equation 1: )$14. Fatigue and fracture of pyrolytic carbon R.O. Ritchie crack! fen (fi. w) ! e260 Pigure 5: (a) Optical micrograph of « Vicker’s indent (inden: tation lout Py = 10 kg) on pyrolytc-carbon coating, showing median radial cracks enuuating from the indent; (b) schematic illustration of the indentation system ane crack configuration showing the characteristic crack-size dimensions a and (37). Measurement Several standards are available for the measurement of fracture toughness values; ASTM Standard E-399 (28) is perhaps the most important, although similar to | Heatt Valve Dis Vol.5. Suppl. 1 corresponding European and Japanese standards, it was developed primarily for metallic materials under solely linear-clastic conditions. Where more than limit- ce inelasticity is present, or where extensive slow crack growth precedes unstable fracture, alternative standards for, respectively, non-linear elastic fracture mechanics (e.g. ASTM E-813) and, resistance-curve behavior (e.g. ASTM E-561) can be used (28). Insimple terms, fracture toughness K, values are iele- ally measured on compact-tension or single-edge- notched bend samples (Fig. 4a), which have been pre cracked (typically by cyclic fatigue) prior to testing; specimens are loaded! monotonically to instability, and the critical value of K computed from the load, P, and crack length, a, at failure. For most standard specimen geometries, handbook K solutions are used (29), eg, for the compact-tension C(1) specimen: «= (raw)! (Wi) . where B and W are respectively the test piece thickness and wielth, andl f(a/W) is a function of a/W tabu- lated in references #28 and #29. Although the toughness is generally clefined under linear-elastic conditions at the onset of unstable fracture, in many toughened ceramics and in metals where extensive plasticity accompanies fracture, slow (table) crack extension, Aa, can occur under increasing stress-intensity levels prior to instability; such behavior is termed a resistance (or R-) curve, and can be used to define the toughness, either at crack initiation (where Aa —> 0) or further out on the R-curve under nominal steady-state conditions (30) Since it is generally difficult to fatigue pre-crack many brittle materials, the fracture toughness of ceramics can be approximated using indentation techniques (31,32). Specifically, the K, value can be estimal- ced from the size of the radial cracks emanating from Vicker's hardness indents at high applied indentation loads (Fig. 5); for pyrolytic carbon, this load is typical ly in excess of 9 kg. The radial cracking forms response to the residval tensile stress field which sur- rounds the indent on unloading; the stress-intensity factor, Ky, from this field can be computed in terms of the peak indentation load Py, and the half surface crack length ¢, by (33): Kge pPine ® where x is a material constant dependent upon the ratio of Young's modulus E, to hardness H,, given by GW: a = BCE)", °J Heart Valve Dis Vol3 Suppl. A ee a 3 tS Boo Mee 2 teme a 2.0 Nien 3 tet eK eahet 13 x BE 2 teh 3 = — 3 510 Ps eg ] Se 7 i a ics Freche Tonto Qi Cee ae ala 00 05 10 1s Crack Extension, a (mm) Figure 6: Resistance curves for half round compact (lenlet) specimens of pyrolytic-carbon coated graphite containing “long” through-thickness (edge) cracks, tester in 37°C Ringer's solution. The fracture toughness, K, cam be defined either af crack initiation, i.c.,a8 Aa 90, or at nominal “steady state” on the plateau of the R-curce, if one exists (37). where 6, is a material-independent constant for Vick- ex’s-produced radial cracks with a value of 0.016 (32). ‘The toughness is then simply computed from Equa~ tions 4 and 5 at K, = Kg by denoting cy as the equilib ‘um surface length of the post-indentation crack (31,32). ‘An example of such radial cracking surrounding, a Vicker’s indent in pyrolytic carbon is shown in Figure 5. Indentation toughness values must be considered to be approximate a3 they take no account of R-curve behavior; in essence, they are defined for crack arrest at some arbitrary point on the R-cury It should be noted that both the ASTM and indentation techniques, described above for measuring fracture toughness values, utilize sharp (ic, fatigue) pre- cracks as their initial stress concentrator. However, certain authors in past have used notched samples to measure the toughness of brittle materials; since the toughness is known to bea strong function of the root radius of the stress concentrator (35); this can lead to unrealistically high K. values, as first pointed out for pyrolytic carbon by this author (25), For example, More et al. (36) used machined-notched double-torsion spec- mens to measure the toughness of pyrolytic carbon and obtained a value of 2.79 « 0,23 MPaYm, some 50% higher than the accepted K, value measured with both indentation (26,37) and pre-cracked (25,37-39) samples. Fatigue and fracture of pyrolytic carbon S15 R.O. Ritchie Toughness of pyrolytic carbon ‘The first fracture toughness measurements on the Si- alloyed LTI pyrolytic carbon-coated graphite material (for the remainder of this paper, the term “pyrolytic carbon” will imply the Sialloyed low temperature isotropic material that is currently used in the manu- facture of heart valve prostheses) using appropriate pre-cracked compact-tension samples yielded K_ val- tues between 1.1 and 1.6 MPavm for tests in room air and 37°C Ringer's solution (25). The latter solution is used here as a simulated physiologic (blood analog) environment and is prepared by dissolving 8.5 g NaCl, 250 mg KCI and 300 mg CaCl, in 1000 ml of distilled water. (There is no difference between results in air and Ringer's solution as little effect of the environment would be expected during the brief duration of these tests). Subsequent studies (13,26,37-39) on both pure (monolithic) pyrolytic carbon and the pyrolytic carbon/graphite laminate generally show complex. R- curve behavior with initiation and steady-state toughness values varying between 1 and 2 MPavm. Within this range, results for pure pyrolytic carbon are invariably at the low end of this range (38,39); for the laminate, the toughness appears to be inereased as the rat of the thickness of the pyrolytic earbon to graphite layers is made smaller (39). It is interesting, to note that with thickness ratios below about 2, the toughness of the composite can be higher than either the pyrolytic carbon or graphite layers. Although not completely Indentation Fracture Toughness ] Pyroyie~Carbon Costing j OE 8 ppc (kg/mm Figure 7: Variation in fracture toughness, K, in the pres: ence of “small” cracks in the pyrolytic-carbon conting as a {function of the indentation fou and equilibriurn radial ‘rack depth P),c,, as mensured using indentation- foughness techniques (37) Frocture Toughness, K, (MPa m'” 60$16 Fatigue and fracture of pyrolytic carbon R.O. Ritchie understood, this has been attributed to variations in the stress intensity across the crack front due to the inho- mogeneous (layered) nature of the material, which ‘causes the crack in the pyrolytic carbon-coating to lead slightly the crack in the graphite substrate (39) ‘An example of fracture-toughness data for pyrolylic- carbon coated graphite is shown in Figure 6; the average K, value, measured in 37°C Ringer's solution at crack initiation on compact specimens machined from actual heart valve leaflets, is 1.64 MPavm (+ 0.53 MPavm) (37). Although there is considerable scatter inherent in the data, the results are consistent with those measured using indentation techniques (Fig, 7), where an average K, of 1.84 MPaVm (+ 0.18 MPaim) has been reported (37). Whereas fracture toughness values between 1 and 2 MPaym are now accepted for monolithic and composite pyrolytic carbon materials, several points are wor- thy of note from these results, First as mentioned above, the K, values quoted aresig- nificantly less than specific results reported! in the litera~ ture based on tests on machined-notched samples (6), Second, the quoted average K, values were defined at crack initiation; for rising R-curves, this value will not be the maximum stress intensity required for crack extension and thus may be expected to be lower than that measured by indentation techniques. Third, the ‘long’ cracks in the compact specimens in Figure 6 are through-thickness (so-called edge cracks) and as such provide an evaluation of the toughness of the pyrocarbon/graphite composite; results from the indentation tests (Fig, 7), on the other hand, sample ‘small’ surface cracks growing principally in the pyrolytic-carbon coating. Fourth, residual stresses, which are known to exist in the composite as a result of coating deposition and processing (e.g., from the thermal-expansion mismatch between the pyrolytic-carbon coating and the graphite substrate), may be expected to affect differently the driving force for ‘small’ cracks in the pyrolytic coating, which is in a state of residual tension, and for ‘long’ cracks which experience a more complex through- thickness variation in residual stress. Fifth, the exact value of the constant &, in Equations 4 and 5 and its material independence, which may affect the magnitude of K, (from Equ. 5), has not been adequately established and remains the subject of some controversy in the literature (32). Moreover, the possible relaxation ofthe residual stresses due to recovery of the pyrolytic carbon may lead to further uncertainties in Equation 4 Finally, recent finite-element calculations (9), which. take into account the composite elastic-property mis- ‘match between the graphite and pyrolytic-carbon layers, indicate some degree of uncertainty in the stress- | Heart Valve Dis Wol.5. Suppl. intensity solutions for the pyrocarbon/sgraphite laminate material. Although no studies to date have taken account of this effect, the numerical calculations do show that the K-solutions for homogeneous material (Equ. 3) may underestimate K in the pyrolytic carbon by up to 40% and overestimate K in the graphite layer by up to 60%. However, as shown below, compa of the fatigue-crack growth properties in the monolithic pyrolytic carbon and the pyrolytic carbon/ graphite laminate, which are characterized in terms of K, do not show stuch major differences (25,37-39) Subcritical crack growth Concepts Although unstable fracture occurs at the fracture toughness value (under plane-strain conditions where K = K), subcritical crack growth can occur at lower K levels by mechanisms such as eyelic fatigue and stress- corrosion cracking (static fatigue) (40); in ceramic materials, this typically occurs at stress intensities above =50% of K, (41). Under cyclic loading conditions, the rate of crack growth per eycle, a/N, can generally be described in terms of power-law relationships, which in their simplest form, can be written as: ta = yy : aN CAK™, (a) & IN =O Krna (AK, (6b) where AKis theapplied stress-intensity range, given by the difference in the maximum and minimum stress- intensity values in the cycle, i.e, AK = Kyoy - Kye © and C’ (= C(1--R)) are experimentally-determined scaling constants, and m = (5 + p) are the crack-growth ‘exponents. Typically, the exponent m takes values in metals between 2 and 4 Gwith p > s), whereas in ceramic materials, m values can be as high as 100 or more (with p << s) (41,42). Similarly, under sustained loading, the corresponding relationships for stress-corrosion crack velocities, a/tt, are of the form: # cry, o where C” and 11 are scaling constants; values of 1 are similar to m, ie, low in metals and high in ceramics. At very low stress intensities, cracks (of a size larger than the dimensions of the microstructure oF scale of local inelasticity) may appear to be dormant; these stress intensities are referred to as the fatigue and stress-cor-| Heart Valve Dis Vol 5. Suppl. 1 rosion thresholds, AKj,, and Ky», respectively. The lic fatigue threshold, AK;,,,, however, is a marked function of the value of the mean (in addition to the alternating) stress, which is generally varied in terms of the so-called load ratio, R = Kyyig/Kyyyy (43). Measterement Traditional cyclic fatigue-crack growth rates are gen erally determined using identical specimen geometries to that used for fracture toughness measurement; this involves monitoring, the growth of through-thickness (edge) cracks, termed ‘long’ cracks where lengths exceed several millimeters, in samples such as the compact-tension specimen (Fig, 8a). Material used for these specimens should accurately reflect the material used in the prostheses, ie. be of clinical grace. Test speci- ‘mens are typically pre-cracked! and cycled, uncer load, displacement or stress-intensity control, on servo- hydraulic testing machines in appropriate environ: ments, eg., in simulated physiologic environments such as 37°C Ringer's solution for heart valve mater als, and the rate of crack growth computed from measurements of the crack length as a function of the num ber of cycles (Fig, 8). ASTM standards have been developed for such measurements on metallic materials, specifically E-647 (28); however, as no standards are available for ceramics, the reader is referred to ref= cerences #25 and #44 fora description of how such tests can be carried out in ceramic-like materials such as pyrolytic carbon, Although standards also do not exist for stress-corrosion testing, crack velocities are similarly obtained for through-thickness ‘long’ cracks in appropriate environments on pre-cracked fracture-mechanics type specimen geometries (Fig. 4a), only under sustained (non-cyclic) loads. For brittle materials like pyrolytic carbon, this is best achieved under load-relaxation conditions, i., under a constant displacement where the load decreases with increasing crack length. Similar to cyclic fatigue, crack velocity /stress intensity (o/K) plots are derived by continuous monitoring of the crack Tength ata known load as a function of time. Cyclic fatigue of pyrolytic carbon Until fairly recently, it had been assumed that tur bostratie carbons such as pyrolytic carbon were insen- sitive to cyclic fatigue degradation. Early studies (18,19) had claimed that the fatigue endurance strength ‘of these materials was virtually identical to the single- cycle fracture stress, i, that cyclic stresses less than this stress do not cause microscopic damage. However, by specifically employing fracture-mechanics type test ing procedures with pre-cracked samples, Ritchie eta 25) first demonstrated unequivocally that fatigue cracks can grow under alternating loads in the pyrolyt- Fatigue and fracture of pyrolytic carbon $17 RO. Ritchie fo) (b) da 09 SN log ak (c) Figure 8: Procedures use for measuring fatigue crack growth rates, showing compact-fension specinne stressed under ‘eyelic lows, AP, b) crack length a versus mumber of cycles, N, curve differentiated to give growth rate (daldN),;at a particu lar crack length ai, and resulting log-log plot of dafaN versus ‘the stress-intensity range, AK. ic-carbon coated graphite Jaminate in both ambient temperature airand 37°C Ringer's solution Subsequent studies have verified that the pyrocarbon/graphite material is indeed susceptible to oyclic fatigue (26,37,3945), and further shown that faligue-crack growth can similarly occur in monolithic$18 Fatigue and fracture of pyrolytic carbon R.O. Ritchie OK thsi" 405: 1 23 10 10" [0% ae 0" ro wl 2 os 2 3. 4 7 STRESS INTENSITY RANGE, BK (MPa-m*) Figure 9: Experimental data shoroing cyetic fatigue-crack propagation rates, dajdN, asa function of the applied stress~ intensity range, AK (= 0.9 Ky), for half rowd compact (leaflet) specimens of pyrotytiexcarbon coated graphite con taining “long” through: thickness (edie) cracks. Results from campact-tension C(T) specimens are included for comparison (25), Tests wore performed in a sinaulated physiological environment of Ringer's solution at 37°C (26). pyrolytic carbon (3839.46). Typical erack-growth rate data, for leaflet specimens ‘of the pyrolytic-carbon/graphite laminate tested in 37°C Ringer’s solution, are shown in Figure 9 asa function of the stress-intensity range, AK (= 09 Kya.) tis apparent that similar to many ceramic materials (41), the slope of these plots, which is characterized by the exponent min Equation 6, is extremely high; values for pyrolytic carbon materials are routinely well over 100 67-39). Moreover, similar to R-curve data for pyrolytic carbons (Fig 6), the growth-rate behavior shows a very high degree of scatter. Based on a comparison of similar fatigue data for ceramic materials tested on the same mechanicatesting machine (41,42), such a wide difference in growth rates (ata given stress intensity) is Unusual and appears to result from inherent material differences ativibutable to variation in_fabrication- induced residual stresses or local microstructure, Although measurements on the pyrolytie-carbon coating of leaflets used for manufacturing heart valves do in fact display large variations in residual stress from J Meat Valve Di Vol 5. Suppl Constant Kinax |, Graptite/Pyrolyic Kmax = 136 Maen" woes ‘Garton Composite a. EE yf a ° ‘00 2000 700 “000 ‘Time (see) igure 10: The effect of sustained-lond vs. eyelic loading corndi- tions, ata constant Kyyy 0 the subcritical cvack groveth nt pyrolytic carbon coated graphite tested in Ringer's solution at 37°C (blood analog). Note how erack-grozth rates rider ‘eyelie toning (region a) far exeeud those reasired under sustained long (region b) (25) sample to sample (13), which would indeed affect properties, the precise origin of the scatter in toughness and crack-growth data for this material is still uncertain. Similar to behavior in many metals and ceramics 41,47), alow growth rates approaching, 10! m/cycle, an apparent fatigue threshold stress-intensity range, AK,» can be defined, below which the growth of ‘long’ cracks are presumed dormant; in most ceramics, this threshold is of the order of 50% of K_(41). Based on the results from several studies in 37°C Ringer's solution (13,25,37-39), measured values of AK, for the monolithic and composite pyrolytic carbon materials vary between =07 and 2 MPaim, with a typical average value being = MPavm. As with toughness data though, within this range, AKyy values for pure pyrolytic carbon are at the low end (generally <1 MPavm) (38,39), whereas in the laminate, thresholds appear to increase with a decreasing ratio of the thickness of the pyrolytic carbon to graphite layers (39). While the data in Figure 9 apparently show a clear cyclic effect, in view of past scepticism over cyclic fatigue in turbostratic carbons (18,19), and in fact in ceramic materials in general (48), it has been necessary to demonstrate unequivocally that the crack growth is cyclically incuced and not simply a consequence of stress-corrosion cracking at maximum load. To achieve this, crack extension has been monitored with (a) the stress intensity eyclically varied between Kya, aNd Kg| Hoar Valve Dis Vol 5. Suppl. 1 tsi") or oe ce 10 10". J 10” i 10" 10", eed 10° 3 z ro = 10% 3 0 = 10) j 5 3 “ ws 0’ 3 7 ha bo : + Gaphie/ryc Lo’ Ringer's Solutior i! Gasny Caton wet) . (Minnear.et all EI ~ Karbon, ti i (Richter.et al) a rote Gopi, Ut o[__"WostinsontNedoau 10" 01 ore Be STRESS INTENSITY, K(MPa-m') Figure 11: Stress-corrosion crack-gro%th behavior of LTT yrolytic-carbon coated graphite composite tested in 37°C Ringer's solution (25), shout in comparison with data for _glassy carbon teste i ceater (49) and air (50) ant for pyrolytic graphite in ambient temperature air (51). and (b) the stress intensity held constant at the same vale of Kiya A typical result, for Kyyy, = 136 MPavm and Ky = 0.4 MPa, is shown in Figure 10 (25). Itis apparent that, whereas crack extension proceeds readily under cyclic loaciing conditions (region (a), ‘upon removal of the cyclic component by holding at the same Ky, (region (b)), crack-growth rates are markedly reduced, Clearly, similar to behavior reported for other ceramic-like materials (41), a true cyclic fatigue effect is apparent; furthermore, subcritical fatigue crack-growth rates under cyclic loading appear to be far in excess of those under equivalent sustained loading, For these reasons, cyclic fatigue must be considered as the dominant mechanism for the extension of cracks in pyrolytic-carbon heart valves; correspond ingly, growth-rate data for this mechanism provicle the basis for damage-tolerant procedures to predict the safe life of these prostheses, Stress-corrosion cracking in pyrolytic carbon Similar to other ceramics, the various forms of car Fatigue and fracture of pyrolytic carton $19 RO. Ritchie bon are prone to subcritical erack growth under the synergistic action of an applied! load and a moist (or corrosive) environment (25,38,49-52). Such stress-corrosion crack-growth behavior is plotted in terms of the «crack velocity with respect to time, da/dt, asa function of the applied stress intensity, K, in Figure 1 for several types of carbon (25). The data include a LI pyrolytic carbon/graphite composite in an environment of {inger’s solution at a temperature of 37°C (25), glassy carbon in water (49) and in air (60), and data for a pyrolytic graphite in air (1). Crack velocities span many orders of magnitude from 10° to 10 m/sec and, similar to cyclic fatigue-crack growth, show a marked sensitivity to the stress intensity; values of the exponent n Equ. 7) typically exceed 100 for pyrolytic carbon (13), Similar to cyclic fatigue, a threshold stress inten: ty, Kyyy can be defined for crack velocities approaching 10"'m/see; values range between =08 and 18 MPa, with a typical average value of =1.25 MPavm (13,25). ‘Although crack growth by stress corrosion tends to be slower than by eyclic fatigue (Fig, 10), it is still a viable mechanism for the extension of pre-existing defects in pyrolytic carbons; accordingly, it is important to characterize the stress-corrosion behavior in order to permit the reliable clesign of a prosthetic device ‘Thneshold stress intensities Despite the apparent convenience of a threshold for no apparent crack growth in engineering analysis, which essentially implies that one can clesign for infi- nite life, the use of thresholds for either stress corrosion or cyclic fatigue in pyrolytic carbon is not recommend ed for the design and life prediction in prosthetic heart valves. The primary reason for this is that such devices are expected to endure extremely large lifetimes in excess 10° cycles; meaning. that subcritical crack growth, if it takes place, may be occurring (at least initially) at vanishingly small growth rates (e.g., below 10°? m/sec), which may be too difficult or time-con- suming to measure reliably. Moreover, threshold val- ‘ues may be questionable in view of the small-crack effect, as described below. Accordingly, for conservative lifetime prediction of prosthetic heart valves, itis recommended that, where very low growth-rate data cannot be obtained, a linear extrapolation of higher crack growth vs. K data be used; however, itis impera- Live for this extrapolation that experimental data are collected at least below growth rates of =10" m/cycle in fatigue and =10* m/sec in stress corrosion, ‘Small cracks Asdescribed above, most fatigue data for heart valve implants are derived from conventional. fracture- mechanics style samples containing ‘long’ (larger than$20. Fatigue and fracture of pyrolytic carbon R.O. Ritchie ‘Small Crack Data a Pyrolytic-Carbon Coating 10" Kae = Kop Kar = Koy + Ky 10° x % é 1 t : aw 10° 10” 10° Surface Crack-Growth Rate, da/dN (mlcycle) (a) Kou (MPa m'*) ‘Small and Long Crack Data ‘3t798 specimen th Sess oomen ie 10 1 Koon = Keg + Ky(MPa m'") Figure 12: Experimental data showing eylicfatigue-crack propagation rates, dala, in pyrolytie 9 kg), act to facilitate fatigue-crack initiation (Fig. 5). Typically, an increasing bending stress range, Aq}, (+10 to 80 MPa) is applied until one crack begins to grow, whereupon the stress is held constant until subsequent crack arrest; the procedure is then repeated. Tests are periodically interrupted every 102 to 10! cycles to permit the use of optical microscopy or gold-coated cellulose-acetate replicas to monitor crack growth, Full details are given in reference #37. For the indentation crack configuration, there are two contributions to the stress intensity, namely that due to the applied stresses and that due to the residual stress field surrouncling the indent, ie, K= Kyo + Ky ‘The stress-intensity factor resulting, from the residual stress, Ky, can be calculated from the peak indentation Toad, Pyyy and the half surface-crack length, c, using,| Heart Valve Dis Vol. 5: Suppl Equations 4 and 5. Linear-elastic solutions for the applied Ky, are given in reference #57 for three- onal semi-elliptical surface cracks in bending, (and /or tension) in terms of the crack depth a, surface ‘crack length 2c, elliptical parametric angle 6, shape factor Q, specimen thickness f, specimen width b, and remole (outer surface) bending stress oy from the applied load P,,,, (Fig. 4b) dimen aa! k-110,(5) Faale.aft.clb,®), (8) where H, is the bencling multiplier and P is the bound- ary correction factor, the form and functional depen- dencies of which are detailed in reference #57. Equa- tion Bis valid for 0.2 ponent of the valve, namely the pyrolytic-carbon housing, will be examined; in reality of course, all components including the leaflets, etc,, would be similarly analyzed. Stress and stress-intensity analyses, The life-prediction analysis must be focused on the critical and most highly stressed locations, which can be defined by practical experience, ie., where cracks have formed on failed or explanted valves, or more generally on the basis of a stress analysis of the component. Due to the complex geometry of most heart valves, the latter invariably must be performed using, finite-element analysis and should treat both the quasi static and impact (dynamic) stresses developed in the component. For the present application, a schematic illustration Of the cross section of the 29 mm mitral valve housing, is shown in Figure 15. In this case, the housing is made from pure pyrolytic carbon without any graphite substrate; the groove in the cross-section is a seating lip for a metallic stellite stiffening ring, Based on numeri cal calculations under both quasi-static and dynamic/impact conditions, ie., 180 mmHg (24 kPa) transvalvular pressure with a leaflet tip velocity of 5 m/s (13), two critical locations for examination can be defined; Location I (Fig. 15a) is the region of highest (dynamic) stresses on the inside surface of the housing in the seating lip area, where the maximum surface siress is 5.5 MPa (calculated for quasi-static conditions) and 30.1 MPa (calculated for worst-case dynamic conditions), and Location Il (Fig. 15b) is the highest stressed region on the outside surface of the housing at the upper corner of the stiffening ring groove, where the maximum surface stress is 43.3 MPa (calculated for dynamic conditions), assuming a corner root radius of 0.07 mm. In addition to the externally applied stresses, residual stresses pre-existing in the housing must also be considered. Based on the hole-drilling tests (13) on pyrolytic-carbon coated graphite valve leaflets, the residual stresses were found to be approximately uniform within the coating with a value of oy = 34 + 14 MPa. Accordingly, for calculation purposes, a uniform residual stress of oy = 48 MPa is taken asa conservative estimate for the monolithic pyrolytic-carbon housing material. For both critical locations on the housing, worst-case, through-thiekness, edge cracks, which penetrate the free surface, are assumed to pre-exist and to follow a path dictated by the distribution of maximum tensile stress and minimum cross-sectional dimensions. An ‘overestimate of the mode I slress-intensity factor for these cracking configurations is given by (13):| Heart Valve Dis Vol. 5. Suppl. K =(ru)' IE, (@/BS,, + F,G0M)S,1 (10) where 9, +'/0,-0,)*(a) +0, (0A) {0,-9,) (bla) rt = (2-1 (32))" x [9752 +2. 02(a/b) + 0.3701 - sin {ra/2iy) 0s (a /26) Fait) =(3))" [0.923 + 0.19911 ~ sin (na /26))4 cos (nz/26) where, with reference to Figure 15, isthe crack length, bis the minimum cross-sectional width of the housing along the crack path, «, is the maximum tensile stress at the surface origin (i, the maximum principal stress in the orifice), and 6, (a) is the maximum tensile stress at the crack tip. The dimension b is 0.91 mm for Loca- tion J and 0.41 mm for Location Il. Note that Equation Woverestimates the stress intensity, especially at small crack sizes where a/b ->0 (13) In addition, forthe lifetime analyses described below, 0, canbe omitted asthe residual stress pre-exists in both the mechanical property test specimens and the valve components. Subcritical rack growth by cyclic fatigue Al existing cracks in the housing are assumed to propagate subcritically from the onset of loading in physiologic service; crack extension by cyclic fatigue is considered to occur with a loading frequency of 1.2 Hz (equivalent to 72 heart beats per minute or 38 x 10° cycles per year) with no incubation or crack initiation time. Based on cyclic fatigue experiments on pure pyrolytic carbon dise-shapec! compact specimens tested in Ringer's solution at 37°C, the worst-case subcritical crack-growth rates can be described by the relationships (13,38): da/aN = C(AK" = 4.15 x 10% AK)? units: m/cycle. (MPaimy"], (1) For the reasons described above, the existence of a threshold for no crack advance is discarded in the present analysis in order to be conservative; crack growth is thus permitted to occur at vanishingly small stress intensities, according to Equation I. This relationship Fatigue ancd fracture of pyrotytic carbon $25 RO. Ritchie was determined for a loac! ratio of close to zero, repre- sentative of the applied “zero-tension’ loading seen in vivo. It is recognized that the existence of residual stresses in the pyrolytic carbon will change this load ratio locally. However, this is not explicitly considered, in the analysis since these stresses also exist in the pyrocarbon test samples. ‘| By integrating Equation 11 between the limits of a, and a,, the number of loading cycles N, required to cause catastrophic failure by cyclic fatigue is thus given Lym i ee C1), FGI, + FS, ‘The critical crack sizes, a,, at which catastrophic failure of the housing occurs, are computed from Equation 2 using the value of K, for pure pyrolytic carbon. In this case, resistance-curve results for fatigue-precracked, disc'shaped compact-tension samples, performed in 37°C Ringer's solution, yielded K, values between 0.9 and 1.04 MPavm (38). (Note that k, values for pure pyrolytic carbon tend to be lower than those for the pyrocarbon/ graphite composite). In light of published results (25,26,37-39), an estimate of K, = 1.0 MPavm ‘was taken. Itshould be noted that because of the math- ematical nature of the analysis, the precise value of K, (in the range 0.8 to 2.5 MPavm) has little effect on the numerical results Integration of Equation 12 using numerical proce- Table I: Predicted lifetime for cyclic fatigue ofthe housing as a function of instal crack size, a,, wnder worst-case physiological loading. InitialCrack ‘Projected Cycles Lifetime Size,a, to Failure, Ny um) (cycles) (years) Location 6 sactot 12x18 66 27x10 75.6 6 56x108 156 nn 23x10? ‘8 months 128 a 0 Location I 43 14x10" 38x10 45 45x10? 125 a 53x 108 15 51 11x 107 4 months 80 0 0 “assuming 38 x 108 cycles per year.$26 Fatigue and fracture of pyrolytic carbon R.O. Ritchie Proje Lift (years) Inga Crack Sie, a (1) Figure 16: Damage-tolerant lifetime predictions ofthe mini- mute life of the ave housing asa function of the initial defect size, a, pre-existing in the component. Predictions pertain fo ‘worst-case through-thickness edge eracks in mininnon dimen sion 29 mum mitral bilealetvaloes, subjected to continous penk physiologic loading conditions at Location Lon the inside of the housing and Location If on the outsite. dures yields estimates of the number of cycles to fail- ture, and hence life of the housing, under worst-case conditions as a function of the defect size a, pre-existing in the housing. The resulting predictions are plotted schematically in Figure 16; specific numerical results for bath cracking locations are listed in Table IL It is apparent that any pre-existing edge crack in the housing of a length in excess of a, ~ 66 jm inside the orifice at the seating lip Location land =45 4m outside the orifice at Location II can grow by cyclic fatigue and cause failure within 10° cycles, or within 100 years or so, The critical crack size at housing failure is between 41,80 and 128 jim depencting upon the location. Corresponding analyses for the leaflets on this valve yield detectable and critical edige crack sizes which are approximately a factor of two larger (13). Thus, provided NDE procedures can reliably detect all edge cracks in excess of a length of 45 jum in the housing prior to the valve entering service, the most conserva~ tive estimates of the damage-tolerant life in Table II imply that this component is capable of remaining operational under continuous peak in vivo loading for periods exceeding patient lifetimes. Subecritical crack growth by stress-corrosion eracking ‘A similar worst-case analysis can be performed for subcritical crack-growth by stress-corrosion cracking, Based on sustained-load tests on pure pyrolytic carbon, dise-shaped compact specimens tested in Ringer's, J Heart Valve Ds Vol.5. Suppl. solution at 37°C, the worst-case subcritical crack: growth rates are given by (13,38): daft = C"™(K)" =1.36 x 107K) Tunits: m/sec. (MPavim)"]. (13) Note again that the use of Equation 13 is conservative in that no crack-growth threshold is assumed; cracks can thus continue to grow at vanishingly small Klevels, ‘As with cyclic fatigue, by integrating the crack- grovith equation between the limits of initial to final crack size, the life of the housing Lis given by the time f required for an edge crack to propagate from a, and 1 lg da L : aK" mk: I" da «aay C7, TF AIDS, + Fy alO05, Pm 072 Resulls showing the life, L, asa function ofthe pre-existing defect size, a,, based on numerical integration of Equation 14, are sted in Table Il for exige cracks both inside (Location 1) and outside (Location ID the housing. It is apparent that any pre-existing edge crack in the housing of a length in excess of a, ~ 82 um on the inside of the orifice (Location D and =55 yim on theoutside of the orifice (Location ID can potentially grow by stress corrosion andl cause failure of the housing within 100 years or 50, Since these estimates represent larger crack sizes than will cause failure by cyclic fatigue (Table ID, the cyclic fatigue process is clearly rate limiting with respect tothe service life ofthe housing such that the required minimum detectable crack size remains as 45 ua. Table Il: Predicted lifetime forthe stress corrosion of the housing as. function of inital crack size, a, under worst-owse physiological loading. Initial Crack Size, a, Lifetime, 1 Gu (years) Location I: a4 12x10 58 68x10 B 1.2 105 2 96 ot 15 128 0 Location I 25 35x10 48 1.9% 10° 55 140 a7 15 80 0| Heart Valve Dis Yok 5. Suppl. graphite core 8) through: thickness edge crack Figure 17: Schematic illustrations of (a) a through-thickness edge crack, and (8) a senui- elliptical surface crack in a pyrolytic-carbon coated graphite lel. Consideration of semi-elliptcal surface flaws ‘The damage-tolerant analyses described above are designed to be highly conservative; this is apparent with respect to the presumption of pre-existing defects in the valve component, the choice of minimum dimension valves, continuous elevated physiologic loading, conditions, worst-case cracks in highest stressed locations, K solutions which overestimate the stress-intensity levels, worst-case subcritical erack-growth behav ior induced at the immediate onset of loading in blood-analog environments (.e., with no crack initiation period), and the assumption of the absence of a fatigue threshold for no erack growth. ‘The specific analysis performed in this study can be considered as the most conservative as it pertains to the extension of worst-case through-thickness edge cracks. Based on this analysis (and in practice that for the other ‘components of the valve (13)), we have reasonably concluded that if final quality-control/NDE_ procedures can be designed to find, and correspondingly reject, all housings that contain defects exceeding, 45 jim in size, the heatt valve prosthesis should certainly survive in vivo long enough to ensure the safety of the patient. From an engineering perspective, however, reliable detection of cracks as small as 50 jtm in every valve is a difficult proposition, although to ensure the highest degree of safety, this is what is required. Moreover, many of the cracks seen on explanted and failed valves are not necessarily through-thickness edge cracks which extend across the whole section, but rather semi- elliptical cracks which penetrate from the free surface (Fig, 17). Ibis therefore appropriate additionally to consider the latter kind of defect as the value of K will invariably be lower than that for the corresponcling edge crack of the same depth. ‘The stress-intensity solution fora semi-elliptical surface crack is given in Equation 8; although K values may vary around the crack front, calculations based on b) semi-elipical surface crack Fatigue and fracture of pyrolytic carbon § 27 RO, Ritchie this equation yield initial and final crack sizes that are signi cantly larger than those computed above. For example, to censure lifetimes in excess of 100 years in the present valve ‘under peak in vivo stresses, the required detectable crack size for a. semé-clliptical surface crack is approximately a factor of two larger in depth than that for the edge crack (13). Accord- ingly, from the perspective of designing quality-control procedures to screen new valve prostheses for the presence of defects, itis important to distinguish between the two types of cracks. For the bileaflet valve currently under consideration, this translates into developing quality- control procedures which can reliably find all edge cracks in the housing in exeess of approximately 45 jim and all semi-elliptical surface cracks with depths in excess of approximately 100 jim. Techniques to accom plish this are described below. Quality control: proof testing ‘The detection of cracks with climensions in the tens of ‘microns in small components such as heart valve prostheses poses real challenges for quality control. Although optical and scanning electron microscopy can be used successfully for metallic valves where required detectable crack sizes are several hundreds of microns in size (12), it would be difficult, if not impos- sible, to apply these techniques reliably to pyrocarbon valves because critical defect sizes are significantly smaller, crack-opening displacements are substantially reduced, and most importantly cracks in pyrolytic carbon are often sub-surface and thus invisible to optical and electron microscopy. Moreover, the defect sizes are essentially beyond the reliable detection range of other commercially available NDE techniques, such as ultra- sonics and X-rays. In view of this, by far the best procedure for this application, infact for many ceramic com- through the use of the proof test ciple behind the proof test is to load each component to a pre-determined stress which is higher at each location than the stress that it would experience in service (60). Since the component will fail when K exceeds K,, components which contain cracks larger than the critical erack size a., based on the applied proof stress o,, will not survive the test. Thus, if the component des not fracture during the proof test, the (initia pre-existing crack size in that component cannot exceed a, eg, from Equation 2:$28 Fatigue and fracture of pyrolytic carbon RO, Ritehie 2 © ) : as) is nce} ‘The importance of this is that through the use of Equa- tions 9 and 15, a minimum fatigue lifetime can imme- diately be estimated fora valve that survives a particular magnitude of proof stress. Furthermore, by adjusting the proof stress such that is less than, or equal to, the minimum required detectable crack size, survival of the proof test implies that the specific valve contains no cracks large enough to propagate to failure within a patient lifetime, ie, that the prosthesis is mechanically safe. In the author's opinion, the proof testis the only reliable method currently available with sufficient sensitivity to evaluate quantitatively and non-destructively the pre-existing defect population in newly manufactured pyrolytic-carbon prostheses. However, for the test to perform properly, certain conditions must be ret. First, although the test should be simple to con- duct as it must be performed on each and every valve, nevertheless absolutely essential that the loading, during the proof test accurately simulates the direction and location of the stresses that the prosthesis will experience in vivo; if this proves to be difficult with a single proof test, two or more tests may need to be designed, for example, for both the housing and the leaflets, ‘An example of such a proof test that mimics the in vivo effect of transvalvular pressure on the leaflets in a bileaflet valve isto apply pressure pneumatically to the assembled valve to close the leaflets momentarily. Sec- ‘ond, the magnitude of the proof stress must be high ‘enough to ‘find’ crack sizes as small as the required detectable crack size, yet not too large that a large per~ centage of the valves are destroyed during the test. Moreover, appropriate proof stress levels should be determined for both edge and surface cracks, With most pyrolytic carbon valves, these two requirements can be generally met with proof stresses some five to ten times larger than in vivo stresses. Third, in contrast to lifetime analyses, upper-bound values of the fracture toughness and lower-bound val- tues of the geometry factors in the relevant K solutions ‘must be used in Equations 9 and 15 for a worst-case estimate of a,. Finally, itis critical that the prosthesis is not damaged by the proof test itself, for example, by a pre-existing crack extending subcritically and arresting Without catastrophic failure of the valve. Although such an event is unlikely to be due to the highly brittle nature of pyrolytic carbon, it is nevertheless vital that some method of crack monitoring be employed during the proof test to ensure, at least qualitatively, that subcritical cracking does not occur; mounting an acoustic emission probe on the loading train of the proof test to | Heart Valve Dis Vol.5. Suppl. Figure 18: Scanning electron micrograph ofthe surface ofthe leaflet ofan explanted pyrolytic carbon valve, showing evi- dence of small-scale pitting and the possible early stages of low evel cavitation damage detect stress waves from such cracking is one way to achieve this, Concluding remarks Careful mechanical design and life prediction, which must accurately simulate realistic failure modes, are essential elements for the reliable use of ceramic implants which will be exposed to complex physiologic loading and environmentally-induced degradation in clinical applications; in this regard, the structural design of cardiovascular-assist devices places particularly demanding requirements on the pyrolytic-carbon ‘materials used as components of most heart valve prostheses. Specifically, to avoid patient trauma due to ‘mechanical fracture of the valve, prostheses must be ied to endure fatigue lifetimes in excess of 10° mulated physiologic environments with fail- ture rates probably no worse than one in 100,000. ‘The damage-tolerant procedures described in this article have been specifically developed to meet this need. They are designed to be highly conservative and thus are undoubtedly more costly than traditional design and life-estimation procedures that until recently have been common in this industry. However, in light of the individual tragedy of the potential loss of a life and the extremely high cost of litigation which will inevitably ensue, this is a small price to pay. ‘The approach is not without pitfalls, however. It should be remembered, for example, that the establish- ment of quality control and proof-test procedures to reliably reject all valves containing defects larger than the required detectable crack size will not necessarily assure safe lifetimes if cracks can develop subsequent to these NDE procedures. Should cracks, or even local-| Heart Valve Dis Vol 5. Suppl. ly roughened surfaces, be formed on the valve components due to inappropriate handling and implant procedures, during. catheterization, or from cavitation dlamage in service (Fig, 18), lifetimes could be significantly shorter. It is therefore vital that procedures are established to minimize the possibility of such damage from occurring during handling or in vivo. ‘There is another factor to consider. Although pyrolytic carbon has been used successfully in hundreds of thousands of implanted valves and is clearly the current material of choice for this application, itis important to realize that from a purely engineering, perspective, it is an extremely brittle material in the presence of cracks. Infact, to this author’s knowledge, the pyrolytic-carbon heart valve is one of the few, if not the only, stractural application where a ceramic-like material is used in a safety-critical situation, ie., where material failure will most likely lead to loss of human lie. Furthermore, another inherent property of pyrolytic carbon and other ceramic-like materials is that subcritical crack-growth rates (at low homologous temperatures) are exceedingly sensitive to the stress-intensity level, ie, the crack-growth exponents, m and, in Equations 6 and 7 are invariably inthe range of 10 to 100 ‘or more (41). This means that the life of any pyroly! carbon component will be highly dependent upon the pre-existing defect size and the level of applied stress (specifically in damage-tolerant life-prediction procedures, life is proportional to o™ or 6). As a result, small changes in a, will lead to large changes in project- ce life, as can be seen in Tables IL and Ill Moreover, since the crack-growth exponents can approach 100 in pyrolytic carbon, a factor of two increase in applied stress can result in a reduction in projected life by over 20 orders of magnitude! Since such sensitivity is an inherent feature of low toughness materials such as pyrolytic carbon, this alone is justifi- cation for design requirements with the highest degree of conservatism, as is afforded by the damage-tolerant approach. Finally, it must be stated that the success of pyrolytic carbon as a material for prosthetic heart valves is ‘undeniable; to date only a small number of components (probably less than about 50) in valves from several manufacturers are known to have failed after implant, and thereby to have produced life-threatening complications inpatients. However, despite this record, itis still an undeniable fact that pyrolytic car- bbon is an inherently very brittle material and that mechanical failure in prosthetic devices is totally unacceptable, both socially and financially, particularly where loss of human life is involved. Accordingly, the use of worst-case damage-tolerant analyses must be regarded as an essential requirement for the design and quality control of new and existing ‘ceramic’ heart Fatigue and fracture of pyrolytic carbon $29 R.O. Ritchie valve prostheses in order to provide maximum assur ance of patient safety Acknowledgements The author would particularly Tike to acknowledge Professor Reiner H. Dauskardt, now at Stanford Uni- versity, Professor Robert M. McMeeking at UCSB and Dr. Avrom M. Brendzel at St. Jude Medical; much of the work described was performed in collaboration with them, Additionally, he would like to thank Dr. Jim McNaney for computational assistance, and Paul Schmidt and Ralph Kafesjian at Baxter, Ms. Diane Mac- Culloch, DanJ. Chwirut, and William F. Regnaultat the FDA, Axel Haubold at the Medical Carbon Research Institute, Gene Stobbs at Medical Inc., Brett Demchuk at ATS Medical, and Al Beavan at Carbomedics for many useful discussions. References 1. Boktos JC, Akins R), Shim HS, Haubold AD, Agarue- al NK. Carbon in prosthetic devices. In: Deviney MD, ‘Grady TM (eds). Petroleum derived carbons. Wash ington D.C, American Chemical Society, 1976:257- 265 2 Bokros JC. Carbon biomedical devices. 1977;15:355-371 Schoen FJ. Carbon in heart valve prostheses: founda~ tion and clinical performance. In:Szycher M (ed). Bio- compatible polymers, metals and composites. Lan- caster, Technomic, 1983:239-261 4. Haubold AD, Yapp RA, Bokros JC. Carbons for bio- ‘medical applications. In: Bever MB (ed). Eneyclopedia of materials science and engincering. Vol. 1. Oxford, UK, Pergamon Press/Cambridge, MIT Press, 19865514520 5. Dauskardt RH, Ritchie RO. Pyrolytic carbon coatings. In Hench LL, Wilson J (eds). An introduction to bioceramics. Singapore, World Scientific Publ. Co, 1993:261-279 6. Lamview Al, Puglia E, Allen PA. Strut fracture and dise embolization of Bjirk-Shiley mitral valve prosthesis localization of embolized dise computerized axial tomography. Ann Thorac Surg, 1982;34:192-195 7. Lindblom D, Bjork VO, Semb BKH. Mechanical fail "ure of the BjGrkShiley valve: incidence, clinical pre- sentation, and management. J Thorac Cardiovasc Surg 1986 92:894.907 8. Ostermeyer J, Horstkoite D, Bennet J. The Bjirk-Shi- ley 70° convexo-coneave prosthesis. strut fracture problem. Thorae Cardiovasc Surg, 1987,35.71-77 Lindblom D, Rodriguez L, Bjork VO. Mechanical fail ture of the Bjork Shiley valve. } Thorac Cardiovasc Surg 1989,97:95.97 10. Kelpetko V, Moritz A, Mlczoch J, H. Schuravitzki, Carbon$30 Fatigue and fracture of pyrolytic carbon 1 2 uM 16, 20. a 23, 24, 2, R.O. Ritchie Domanig E, Wolner E. Leaflet fracture in Edwards- Duromedics bileaflet valves, J Thorac Cardiovase Surg 1989;97:90-94 Ritchie RO, Dauskardt RH, Pennisi FJ, On the fractog- raphy of overload, stress corrosion and cyclic fatigue failures in pyrolytic-carbon materials used in prosthetic heart valve devices. J Biomed Mater Res: 1992;26:69-76 Ritchie RO, Lubock P. Fatigue life estimation proce- lures for the endurance ofa cardiac valve prosthesis: stress/life and damage-tolerant analyses. J Biomech Eng, Transactions of ASME 1986;108:153-160 Ritchie RO, MeMeeking RM, Schmidt P, Lam HL. Damage-tolerant life-prediction procedures for the endurance of a pyrolytie-carbon mechanical heart valve prosthesis. | Biomed Mater Res 1996;30:(in ress) CChuvirut DJ, Regnault WP, Fracture mechanics prine ples applied to implant medical devices-a review. Medical Progress through Technology 1988;14:193- 203 Replacement Heart Valve Guidance Document, Washington, DC: Division of Cardiovascular, Respi- ratory, and Neurological Devices, Food and Drug, Administration, US. Department of Health and Human Services, 1994 Kollensky WV. Deformation in pyrolytic graphite. ‘Trans Met Soc AIME 1965;223:830-832 Kotlensky WV, Martens HE. Structural changes accompanying deformation in pyrolytic graphite. J ‘Am Ceram Soe 1965;48:195-138 Schoen FJ. On the fatigue behavior of pyrolytic car~ bon. Carbon 1973;11:413-414 Shim HS. The behavior of isotropic pyrolytic carbons under cyclic loading. Biomaterials and Medical Devices: Artificial Organs 1974;2:55-65 Kaae JL, Gulden TD, Liang S. Transmission electron microscopy of pyrolytic carbons deposited ina bed of fluidized particles, Carbon 1972;10:701-709 Bokros JC, LaGrange LD, Schoen FJ. Control of struc- tureof carbon for use in bioengineering, In: Walker PL (ed). Chemistry and physics of carbon. New York, NY, Dekker, 1972:103-171 Kaae JL. Microstructures of pyrolytic earbon/silicon carbide mixtures co-deposited in a bed of fluidized particles. Carbon 1975;13:51-53 Pollmann E, Pelissier J, Yust CS, Kaae JL. Transmis- sion electron microscopy of pyrocarbon coatings. ‘Nuclear Technol 1977;35:301-309 Kaae JL. The effect of annealing on the microstrue- tures and the mechanical properties of poorly crys- talline isotropic pyrolytic carbons. Carbon 1972;104691-699 Ritchie RO, Dauskardt RH, Yu W, Brendzel AM. Cyclic fatigue-crack propagation, stress-corrosion| m7. 28, 29, 20. 32, 37. 38, 39. | Hart Valve Dis Wol.5. Suppl. and fracture-toughness behavior in pyrolytic carbon coated graphite for prosthetic heart valve applica tions.) Biomed Mater Res 1990;24:189-206 Ritchie RO, Dauskardt RH, Brendzol AM. Role of small cracks inthe structural integrity of pyrolyticear bon heart valve prostheses. In: Ducheyne P, Chris tionsen D (eds). Bioceramics. Vol 6 Proceedings of the 6th International Symposium on Ceramies in Medi- cine), Butterworths-Heinemann, 1993:229-236 Mitchell MR, Fundamentals of moctern fatigue analysis for design. In: Fatigue and microstructure. Metals Park, OH, American Society for Metals, 1979:385-466 Annual book of ASTM standards. Vol. 3.01, Section 301. Philadelphia, PA, American Society for Testing and Materials, 193, ‘Tada H, Paris PC, Irwin GR. Stress analysis of eracks handbook. 2nd Ed. St. Louis, MO, Paris Publica- tions/Del Research Corp., 1985 Lawn BR. Fracture of brittle solids. 2nd! Ed. Cam- bridge, UK, Cambridge University Press, 1993 Lawn BR, Evans AG, Marshall DB. Elastic/plastic indentation damage in ceramics: the median/radial czack system. | Am Ceram Soc 1980;63:574-581 Ponton CB, Rawlings RD. Vickers indentation fracture toughness test. Part 1. Review of literature and formulation of standardized indentation toughness equations. Mater Sci Tech 1989;5:865-872 Lawn BR, Puller ER. Equilibrium penny.-like cracks in inclentation fracture. J Mater Sci 1975;10:2016-2024 ‘Anstis GR, Chantikul P, Lawn BR, Marshall DB. A critical evaluation of indentation techniques for measuring fracture toughness: 1 direct crack measurements. J Am Ceram Soe 1981 64533-5365 Ritchie RO, Francis B, Server WL. Evaluation of toughness in AISI 4340 alloy steel austenitized at low and high temperatures. Metall Trans A 19767A:831- 838 More RB, Haubold AD, Beavan LA. Fracture toughness of Pyrolite® carbon. ‘Trans Soc Biomater 198915180 Dauskardt RH, Ritchie RO, Takemoto JK, Brendzel 'AM. Cyclic fatigue and fracture in pyrolytic carbon- coated! graphite mechanical heart valve prostheses: role of small cracks in life prediction. J Biomed Mater Ros 1994;28791-804 Ritchie RO, Dauskardt RH. Fracture toughness and suberitcal crack: growth behavior of Pyrolit in simulated physiological environments. Technical report to CarboMedics, Inc, November 1980, Cited in: Beavan LA, James DW, Kepner J L. Evaluation of fatigue in Pyrolite® carbon. In: Ducheyne P, Christiansen D (cs). Bioceramics. Vol 6. (Proceedings of the 6th International Symposium on Ceramics in Medicine) Butterworths Heinemann, 1998:205-210 Gilpin CB, Haubold AD, Fly JL. Fatigue crack growth| Hl Valve Dis Vol 5. Suppl 40. a 2, 8, aM, 46. a7. 48, 49, and. fracture of pyrolytic carbon composites. I Ducheyne P, Christiansen D (eds). Bioceramics. Vol 6 (Proceedings ofthe 6th International Symposium on Ceramics in Medicine). Butterworths-Heinemann, 1993:217-223 Johnson HH, Paris PC. Sub-eriical flaw growth, Eng Fract Mech 1968;1:3-45 Ritchie RO, Dauskardt RH. Cyelic fatigue in ceramics a fracture mechanics approach to subcritical crack growth and life prediction, J Ceram Soc Japan 191;99:1047-1062 Dauskardt RH, James MR, Porter JR, Ritchie RO. Cyclic fatigue-crack growth in SiC-whisker-rein forced alumina ceramic composite: long and small- crack behavior. J Am Ceram Soc 1992;75:759-771 Gilbert CJ, Dauskardt RH, Ritchie RO. Behavior of cyclic fatigue cracks in monolithic silicon nitride.) Am ‘Ceram Soc 1995;78:2291-2300 Dauskardt RH, Ritchie RO. Cyclic fatigue-crack growth behavior in ceramies. Closed Loop 1989:17:7- 7 Sines G, Ma Ling. Long life fatigue of pyrolytic carbon. In: Ducheyne P, Christiansen D (eds). Bioceram- ies. Vol 6, (Proceedings ofthe 6th I onal Syn posium on Ceramics in. Medicine) Butterworths-Heinemann, 1993:211-215, Beavan LA, James DW, Kepner JL. Evaluation of fatigue in Pyrolito® carbon. In: Ducheyne P, Chris- tiansen D (eds). Bioceramics. Vol 6. (Proceedings of the th International Symposium on Ceramics in Medicine). Butterworths-Heinemann, 1988:205-210 Ritchie RO. Near-threshold fatigue crack propagation insteols Int Metals Rev 1979;20:205-230 Evans AG. Fatigue in cerarsics. Il J Fract 1980;16:485- 495 Minnear WP, Hollenbeck TM, Bradt RC, Walker PL. Suib-critical erack growth of glassy carbon in water. J Non-Cryst Solids 197621:107-118 Fatigue and fracture of pyrolytic carbon S31 R.O. Ritchie 50, Soltess U, Ritter H. Mechanical behavior of selected ceramics. In: Ducheyne P, Hastings GW (els). Metals, and ceramies biomaterials. Vol 2. Strength and surface. Boca Raton, CRC Press, 198423461 51. Hodkinson PH, Nadeau JS. Slow crack growth in graphite. J Mater Sci 1975;10:846-856 52. Ely JL, Haubold AD. Static fatigue and stress eorro- sian in pyrolyticearbon. In: Ducheyne P, Christiansen D (cals). Bioceramics. Vol 6. (Proceeelings of the 6th International Symposium on Ceramnies in Medicine) Butterworths-Heinemann, 1993:199-204 Suresh §, Ritehio RO. The propagation of short fatigue cxacks. nt Metals Rev 1984;29:445-476 54, Ritchie RO, Lankford J (cs). Smal fatigue cracks Warrendale, PA, The Metallurgical Society of AIME, 1986065 55. Miller KJ cle fos Rios ER (eds). The behaviour of short fatigue cracks. London, UK, Institute Mechanical Engineers, 1986:560 56, Steffen AA, Dauskart RH, Ritehie RO. Cyclic fatigue life and crack-growth behavior of microstructurally- samall cracks in Mg-PSZ. ceramics. J Am Ceram Soe 1991;74:1259-1268 57. Raju IS, Atluri SN, Newman, JC J. Stress-intensity factors for small surface and comer cracks in plates. In: Wei RP, Gangloff RP (eds). Fracture mechanics perspectives and clieetions 20th Symp). ASTM STP 1020. Philadelphia, PA, Am Soc Testing Mater, 1989:297-316 58, Ritchie RO. Mechanisms of fatigue crack propagation Jn metals, composites and ceramics: role of erack-tip shielding. Mater Sci Eng, 1988;103A:15-28 59, Evans AG. The new high toughness ceramics. In: Wei RP, Gangloff RP (eds). Fracture mechanics: perspectives and directions (20th Symp.) ASTM STP 1020 Philadelphia, PA, Am Soe Testing Mater, 1989:267-291 60. Evans AG, Wiederhorn SM. Proof testing of ceramic materials - an analytical basis for failure prediction. IntJ Fract 1974;10379-392

1996 FatigueFracturePyrolyticCarbon Ritchie

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

1996 FatigueFracturePyrolyticCarbon Ritchie

Uploaded by

Copyright:

Available Formats

You might also like